diff --git a/CHANGELOG.md b/CHANGELOG.md
index a6e8ee203..0c55cd318 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,10 @@
 
 ## Unreleased
 
+### Added
+
+- Blog posts for releases 0.43.0 through 0.52.0, including the 0.50.1 cleanup patch
+
 ## [0.52.0](https://github.com/TypedDevs/bashunit/compare/0.51.0...0.52.0) - 2026-10-11
 
 ### Changed
diff --git a/docs/blog/2026-07-24-release-0-43.md b/docs/blog/2026-07-24-release-0-43.md
new file mode 100644
index 000000000..0de2b541d
--- /dev/null
+++ b/docs/blog/2026-07-24-release-0-43.md
@@ -0,0 +1,40 @@
+---
+date: '2026-07-24'
+title: 'Release 0.43.0'
+description: 'See which loops your tests reach, with fewer extra processes during checks and parallel tests.'
+coverAlt: 'Reports of loops reached by tests and fewer extra processes during checks'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.43.0 reports whether your tests enter loops. It also reduces the extra processes started by checks, called assertions, and by tests running at the same time. Existing checks keep their behavior, so you do not need to rewrite tests.
+
+## Find loops your tests never enter
+
+The report now includes `while`, `until`, `for`, and `select` loops. Each loop body counts as one possible path through the code, called a branch. That path is marked reached when the loop runs at least once. A test with only empty input can leave the loop body untested even when the surrounding function runs successfully ([#855](https://github.com/TypedDevs/bashunit/issues/855)).
+
+```bash
+bashunit tests/ --coverage --coverage-paths src/
+```
+
+For a function that processes a list, include both an empty-list case and a case with at least one item. The first checks what happens when the loop is skipped; the second reaches its body. The report now makes that missing test case visible. It records whether the loop was entered, rather than treating each pass through the loop as a separate path.
+
+## Fewer processes for the same checks
+
+Text checks such as `assert_contains`, `assert_matches`, and checks of the start and end of a value no longer start an extra shell process to join their arguments. Checks of arrays, dates, durations, JSON, files, and folders also find their failure labels without starting a shell process to capture a value. The labels and comparisons stay the same ([#844](https://github.com/TypedDevs/bashunit/issues/844)).
+
+Tests running at the same time now name result files using a sequence number assigned by the test run. This replaces creating and renaming a temporary file for every test. Tests supplied with different sets of input values also avoid the chain of commands previously used to make those inputs safe for filenames. Results are still collected in the same way ([#851](https://github.com/TypedDevs/bashunit/issues/851)).
+
+```bash
+bashunit --parallel tests/
+```
+
+These changes reduce repeated work as the number of checks and tests grows. They add no new options and do not change how you run the tests.
+
+---
+
+See the full changelog for [0.43.0](https://github.com/TypedDevs/bashunit/releases/tag/0.43.0) on GitHub.
diff --git a/docs/blog/2026-07-29-release-0-44.md b/docs/blog/2026-07-29-release-0-44.md
new file mode 100644
index 000000000..af7897171
--- /dev/null
+++ b/docs/blog/2026-07-29-release-0-44.md
@@ -0,0 +1,52 @@
+---
+date: '2026-07-29'
+title: 'Release 0.44.0'
+description: 'Test your own checks, manage saved expected output, and catch command-tracking and setup mistakes.'
+coverAlt: 'Testing custom checks, reviewing saved expected output and catching setup mistakes'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.44.0 helps you test and document your own checks, called custom assertions. It also gives you more control over saved expected output and catches setup mistakes that could previously produce misleading results.
+
+## Build and test your own checks
+
+`bashunit::assert_that` runs a command and records whether the check passed in one call. `bashunit::assert_once` lets a check made of several smaller checks count and report once under its own label. Existing totals stay unchanged unless you choose to use it ([#915](https://github.com/TypedDevs/bashunit/issues/915), [#917](https://github.com/TypedDevs/bashunit/issues/917)).
+
+The new `assert_assertion_passes`, `assert_assertion_fails`, and `assert_assertion_fails_with` helpers test what your check reports. They keep its counts and output separate, so an expected failure does not fail the surrounding test ([#916](https://github.com/TypedDevs/bashunit/issues/916)).
+
+```bash
+function test_positive_assertion() {
+  assert_assertion_passes assert_positive_number 3
+  assert_assertion_fails_with "positive number" assert_positive_number 0
+}
+```
+
+Here, `assert_positive_number` is your project's check. Load the setup file that defines it to list its documentation alongside the built-in reference. Use `--custom` to list only your project's checks ([#918](https://github.com/TypedDevs/bashunit/issues/918)).
+
+```bash
+bashunit doc --custom --boot tests/bootstrap.sh
+```
+
+## Review saved expected output
+
+A snapshot is saved expected output that later runs compare against. Use `--snapshot-update` to replace existing snapshots after an intentional change, optionally narrowing the run with `--filter`. Snapshots containing markers for text you chose to ignore are left alone. In automated checks, `--no-snapshot-create` makes a missing snapshot fail instead of silently saving the first reference. `--snapshot-report-unused` lists leftovers from renamed or deleted tests without deleting anything. It requires a full run ([#900](https://github.com/TypedDevs/bashunit/issues/900), [#901](https://github.com/TypedDevs/bashunit/issues/901), [#902](https://github.com/TypedDevs/bashunit/issues/902)).
+
+```bash
+bashunit tests/ --no-snapshot-create
+bashunit tests/ --snapshot-report-unused
+```
+
+## Catch mistakes before trusting a pass
+
+A spy tracks calls to a command. Checks of those calls now reject a command that was never registered for tracking, instead of reporting zero calls. Comparing arguments one at a time distinguishes one argument containing a space from two separate arguments. Failed checks show the recorded calls.
+
+Invalid options, numeric settings, and missing setup or environment files now produce errors. Coverage, which reports how much code the tests reach, also gains an `xtrace` method of recording execution on Bash 4.1 and later. Automatic selection keeps the existing `trap` method on older shells. Bash 3.0 remains the minimum supported version.
+
+---
+
+See the full changelog for [0.44.0](https://github.com/TypedDevs/bashunit/releases/tag/0.44.0) on GitHub.
diff --git a/docs/blog/2026-08-09-release-0-45.md b/docs/blog/2026-08-09-release-0-45.md
new file mode 100644
index 000000000..26746ebb7
--- /dev/null
+++ b/docs/blog/2026-08-09-release-0-45.md
@@ -0,0 +1,50 @@
+---
+date: '2026-08-09'
+title: 'Release 0.45.0'
+description: 'Check symbolic links and commands, save multiple expected outputs, and avoid misleading test results.'
+coverAlt: 'Symbolic-link and command checks, saved expected outputs and corrected results'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.45.0 adds ways to check symbolic links, pass arguments to command checks, and save multiple expected outputs in one test. It also fixes cases where invalid inputs or missing commands could produce misleading results.
+
+## Check the link itself
+
+A symbolic link is a path that points to another file or directory. Most file checks follow the link and check its target. The new `assert_is_symlink`, `assert_is_not_symlink`, and `assert_symlink_to` check the link itself. They are useful for scripts that keep a link pointing to the current release ([#981](https://github.com/TypedDevs/bashunit/issues/981)).
+
+```bash
+function test_current_release_link() {
+  assert_is_symlink "./current"
+  assert_symlink_to "./releases/42" "./current"
+}
+```
+
+## Pass arguments directly to command checks
+
+`assert_true` and `assert_false` now accept a command followed by its arguments. A single argument retains its previous meaning ([#994](https://github.com/TypedDevs/bashunit/issues/994)).
+
+```bash
+function test_release_directory() {
+  assert_true test -d "./releases/42"
+  assert_false test -d "./releases/missing"
+}
+```
+
+Both checks now fail when the command is unknown or cannot be executed. A missing command is no longer accepted by `assert_false` as evidence that the intended condition was false; the failure message explains that the command could not run ([#982](https://github.com/TypedDevs/bashunit/issues/982)).
+
+## Keep expected outputs separate and results reliable
+
+Snapshots save expected output for comparison on later runs. Giving each snapshot a name lets a test save more than one expected value. When values differ, bashunit shows the snapshot file path and the `--snapshot-update` hint. Test paths starting from the filesystem root now find snapshots correctly. When Perl is unavailable, the alternative comparison checks the whole value, including any markers for text you chose to ignore ([#986](https://github.com/TypedDevs/bashunit/issues/986), [#1002](https://github.com/TypedDevs/bashunit/issues/1002)).
+
+Input checks are stricter too. Core comparison checks reject missing arguments, date checks reject values they cannot read as dates, and `assert_json_equals` rejects invalid JSON. Runs with tests executing at the same time retain results from same-named test files in different directories.
+
+Several existing checks start fewer extra processes. These include snapshots without ignored-text markers, counters that track command calls, text searches that ignore letter case, and common comparisons that allow a small numeric difference. The single-file executable also omits source comments while keeping text blocks inside shell commands and markers identifying source files. These changes reduce extra work while preserving how the checks behave.
+
+---
+
+See the full changelog for [0.45.0](https://github.com/TypedDevs/bashunit/releases/tag/0.45.0) on GitHub.
diff --git a/docs/blog/2026-08-11-release-0-46.md b/docs/blog/2026-08-11-release-0-46.md
new file mode 100644
index 000000000..dd5571f67
--- /dev/null
+++ b/docs/blog/2026-08-11-release-0-46.md
@@ -0,0 +1,50 @@
+---
+date: '2026-08-11'
+title: 'Release 0.46.0'
+description: 'Choose tests, find intermittent failures, and report results and changed-line coverage in automated checks.'
+coverAlt: 'Selecting changed tests, finding intermittent failures and reporting changed lines reached by tests'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.46.0 makes it easier to choose tests, investigate failures that happen only sometimes, and publish results from automated checks. It also fixes reports that could contain zero tests after running tests at the same time.
+
+## Inspect and select a run
+
+`--changed` selects test files Git reports as changed since a chosen commit or branch. It includes changes prepared for a commit, other edits, and new files not yet tracked by Git. It selects changed test files, rather than finding every test affected by a source-code change. Use `--list` to inspect the selected tests without executing them ([#1010](https://github.com/TypedDevs/bashunit/issues/1010), [#1007](https://github.com/TypedDevs/bashunit/issues/1007)).
+
+```bash
+bashunit tests/ --changed origin/main --list
+bashunit tests/ --list --list-format json
+```
+
+Tag expressions can require two tags together with AND or leave out a tag with NOT. Repeating `--tag` still selects tests that match any of those options. `--exclude-filter` excludes matching test names even when another option selected them. `--order-by defects` runs the previous failures first while still running every selected test.
+
+## Find tests that fail only sometimes
+
+`--repeat` runs every selected test several times, reports one combined result, and identifies which repetition failed. Tests that pass only after a retry are now counted separately as flaky, with the first failure included in reports. They still count as passed, so the run can succeed. Use `--fail-on-flaky` to make those results fail the run ([#1013](https://github.com/TypedDevs/bashunit/issues/1013), [#1012](https://github.com/TypedDevs/bashunit/issues/1012)).
+
+```bash
+bashunit tests/ --repeat 20 --filter flaky_candidate
+bashunit tests/ --retry 2 --fail-on-flaky
+```
+
+## Publish useful reports from automated checks
+
+JSON, TAP, HTML, and JUnit file reports now collect results correctly under `--parallel`. JUnit groups tests by file, includes their recorded output, and carries informative failure messages. Markdown summaries can appear in the GitHub Actions step summary. Automatic error notices in Actions point to the failing file and line ([#1004](https://github.com/TypedDevs/bashunit/issues/1004), [#1016](https://github.com/TypedDevs/bashunit/issues/1016)).
+
+Coverage reports which lines the tests reach. It gains the Cobertura XML format used by other automated-check tools. `--coverage-diff` limits the report and minimum-percentage check to changed lines:
+
+```bash
+bashunit tests/ --coverage --coverage-diff origin/main --coverage-min 90
+```
+
+Preparing coverage reports also avoids starting an extra process to decide whether each source line can run. New checks test whether a number falls within a range and whether a command is available. Invalid numbers and failures from checks you define now receive clearer handling.
+
+---
+
+See the full changelog for [0.46.0](https://github.com/TypedDevs/bashunit/releases/tag/0.46.0) on GitHub.
diff --git a/docs/blog/2026-08-13-release-0-47.md b/docs/blog/2026-08-13-release-0-47.md
new file mode 100644
index 000000000..52fd3f4a1
--- /dev/null
+++ b/docs/blog/2026-08-13-release-0-47.md
@@ -0,0 +1,48 @@
+---
+date: '2026-08-13'
+title: 'Release 0.47.0'
+description: 'Pipe reports to other commands, reuse test settings, catch unexpected commands, and include untested files in coverage.'
+coverAlt: 'Reports sent to other commands, reusable test settings and coverage including untested files'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.47.0 lets you send reports directly to other commands, reuse settings for groups of tests, and control individual tests. Coverage, which reports how much code the tests reach, now includes tracked files that no test ran. Keep that change in mind when comparing percentages after an upgrade.
+
+## Send reports straight to another command
+
+`--output json` and `--output junit` write a report to the normal terminal output, which you can pipe to another command or redirect to a file. You do not need a temporary report file. The existing options that write reports to files remain available alongside them ([#1018](https://github.com/TypedDevs/bashunit/issues/1018)).
+
+```bash
+bashunit tests/ --output json | jq '.summary'
+bashunit tests/ --output junit > report.xml
+```
+
+Named `[suite:<name>]` sections in `.bashunitrc` store paths and options for groups of tests. Use `--suite <name>` to select one group, repeat the option to combine groups, or use `--list-suites` to see available names ([#1021](https://github.com/TypedDevs/bashunit/issues/1021)).
+
+```bash
+bashunit --list-suites
+bashunit --suite unit
+```
+
+## Give each test the settings it needs
+
+The comments `# @timeout`, `# @retry`, and `# @skip` above a test set its time limit, retries, or skipped status. They override settings for the whole run. Invalid values stop the run instead of quietly using another setting. The new helpers for skipping a test under certain conditions both mark it skipped and end it, replacing a separate `return` ([#1020](https://github.com/TypedDevs/bashunit/issues/1020), [#1019](https://github.com/TypedDevs/bashunit/issues/1019)).
+
+`--sandbox` fails a test that runs an external command it did not replace with a test-controlled response, except commands already allowed by default. `--sandbox-allow` adds more allowed commands. To test code that retries, `bashunit::mock_sequence` supplies a different answer on each call and repeats the final answer after all the listed answers have been used ([#1022](https://github.com/TypedDevs/bashunit/issues/1022), [#1023](https://github.com/TypedDevs/bashunit/issues/1023)).
+
+## Include files no test reached
+
+Previously, coverage considered only files a test ran. Now it includes every file under `--coverage-paths`, with untouched files reported at zero coverage. Percentages may drop because the report includes more of the code. Reports limited to changed lines also include changed files that no test ran. A new untested file therefore cannot escape the minimum-percentage check ([#1053](https://github.com/TypedDevs/bashunit/issues/1053), [#1054](https://github.com/TypedDevs/bashunit/issues/1054)).
+
+Fixes to how bashunit reads alternatives in `case` statements and finds where functions begin and end also change coverage details. Preparing the report now uses one call to `awk` per run to identify executable lines and write the output, rather than repeating shell loops and starting processes for each file.
+
+Runs that measure execution time can now save JSON and JUnit reports and compare the middle timing sample, called the median, against a saved reference. `--snapshot-prune` removes unused snapshots, the files holding saved expected output, only after a full run without failures. It prints each deleted path. Runs that select only some tests cannot remove them ([#1029](https://github.com/TypedDevs/bashunit/issues/1029), [#1030](https://github.com/TypedDevs/bashunit/issues/1030)).
+
+---
+
+See the full changelog for [0.47.0](https://github.com/TypedDevs/bashunit/releases/tag/0.47.0) on GitHub.
diff --git a/docs/blog/2026-08-14-release-0-48.md b/docs/blog/2026-08-14-release-0-48.md
new file mode 100644
index 000000000..3bde3e7ca
--- /dev/null
+++ b/docs/blog/2026-08-14-release-0-48.md
@@ -0,0 +1,38 @@
+---
+date: '2026-08-14'
+title: 'Release 0.48.0'
+description: 'Faster coverage reports, clearer errors and fixes for missing tests and unreadable reports.'
+coverAlt: 'Release 0.48.0 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.48.0 makes coverage reports faster and explains more clearly why a run failed. It also fixes cases where running tests at the same time could report success while silently missing tests.
+
+## Faster coverage and HTML reports
+
+Coverage reports show which lines of code your tests ran. bashunit now does less repeated work when reading those results. In measurements of this repository, a coverage run fell from 16.2 seconds to 2.9 seconds, while creating an HTML report for 128 files fell from 58.7 seconds to 3.1 seconds ([#1092](https://github.com/TypedDevs/bashunit/issues/1092), [#1117](https://github.com/TypedDevs/bashunit/issues/1117)). Other projects will have different timings.
+
+The HTML test report now includes a Failures section with each failed test's name, file, line and message. It also counts risky tests (which may not have checked anything) and flaky tests (which passed only after a retry). Special characters in test titles are written safely so they cannot become HTML code ([#1251](https://github.com/TypedDevs/bashunit/issues/1251), [#1252](https://github.com/TypedDevs/bashunit/issues/1252), [#1249](https://github.com/TypedDevs/bashunit/issues/1249)). Coverage reports also handle filenames containing `|`, `<` or `&`.
+
+On Bash 3.x, `--verbose` explains that coverage cannot count code run in a separate shell process. On supported newer shells, coverage now keeps the results from commands run inside `$(...)` instead of losing them ([#1112](https://github.com/TypedDevs/bashunit/issues/1112), [#1101](https://github.com/TypedDevs/bashunit/issues/1101)).
+
+## Missing tests become visible failures
+
+Repeated test-function names are detected when tests run at the same time. A data provider, the function that supplies inputs for a test, now produces a named error if it is missing or returns no inputs. The test no longer disappears behind “No tests found” ([#1147](https://github.com/TypedDevs/bashunit/issues/1147), [#1145](https://github.com/TypedDevs/bashunit/issues/1145)).
+
+A setup file that fails to load now makes the run fail wherever that file is needed. Trying to write a report to a folder gives a clear error. Asking for a speed measurement where there are no benchmark functions also fails. Automated checks can now catch these problems instead of treating them as successful runs.
+
+## Reports that other tools can read
+
+JUnit reports now contain valid XML. Stopping a parallel run after a failure no longer mixes terminal messages into reports read by other tools. TAP, another test-report format, handles `#` in names so `# SKIP` cannot accidentally make a test look skipped ([#1239](https://github.com/TypedDevs/bashunit/issues/1239), [#1243](https://github.com/TypedDevs/bashunit/issues/1243), [#1119](https://github.com/TypedDevs/bashunit/issues/1119)).
+
+`assert_exec` can now check a command that is expected to fail even with `--strict`. Comparisons preserve literal backslashes, and a supplied test input ending in a backslash reaches the test unchanged. The docs also explain lowercase `test_` names, passing installer options by position, and giving each set of test inputs its own saved expected output.
+
+---
+
+See the full changelog for [0.48.0](https://github.com/TypedDevs/bashunit/releases/tag/0.48.0) on GitHub.
diff --git a/docs/blog/2026-08-16-release-0-49.md b/docs/blog/2026-08-16-release-0-49.md
new file mode 100644
index 000000000..48f26d7d1
--- /dev/null
+++ b/docs/blog/2026-08-16-release-0-49.md
@@ -0,0 +1,47 @@
+---
+date: '2026-08-16'
+title: 'Release 0.49.0'
+description: 'Choose when an empty run is allowed, list available tags and keep large test runs complete.'
+coverAlt: 'Release 0.49.0 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.49.0 distinguishes a run with no matching tests from a path that does not exist. It also lists available tags and fixes problems that could stop a large test run before all its tests had executed.
+
+## Allow a run with no tests deliberately
+
+By default, finding no tests remains a failure. `--pass-with-no-tests` allows that case to succeed while still printing “No tests found”. This helps when automated checks split tests across machines and one machine has none to run, or when none of the changed files contain tests ([#1263](https://github.com/TypedDevs/bashunit/issues/1263)).
+
+```bash
+bashunit --changed origin/main --pass-with-no-tests tests/
+```
+
+A path that does not exist is still an error. An empty folder you explicitly select also no longer makes bashunit run `BASHUNIT_DEFAULT_PATH` instead. The runner cannot report success for a different set of tests from the one you requested.
+
+## Find the tags you can select
+
+`--list-tags` prints the selected files' tags in sorted order, without duplicates, and runs no tests:
+
+```bash
+bashunit --list-tags tests/
+bashunit --tag slow tests/
+```
+
+If a tag matches nothing, bashunit lists the tags it saw, or explains that no test has one. That makes a spelling mistake easier to distinguish from a run with no matching tests ([#1265](https://github.com/TypedDevs/bashunit/issues/1265)).
+
+## Keep large test runs complete
+
+When running files one after another, bashunit now closes the file handles it uses. Previously those handles could reach the system limit, after which tests stopped checking results even though the run still reported success. In a test case with 120 files and a limit of 120 open handles, only two checks ran instead of 120 ([#1271](https://github.com/TypedDevs/bashunit/issues/1271)).
+
+Cleanup after each test also stops searching the entire shared temporary folder. The recorded 100-test example fell from 978 ms to 542 ms with 5,000 leftover files present. Use `bashunit::temp_file` and `bashunit::temp_dir` for automatic cleanup; files written there by hand are no longer removed for you ([#1269](https://github.com/TypedDevs/bashunit/issues/1269)).
+
+Snapshot files store expected test output. The commands that list or remove unused snapshots now find ones left by deleted or renamed test files. They still leave snapshots belonging to existing files outside the current run alone. Review the listed paths before deleting them: this release can remove more unused snapshots than earlier versions ([#1194](https://github.com/TypedDevs/bashunit/issues/1194)).
+
+---
+
+See the full changelog for [0.49.0](https://github.com/TypedDevs/bashunit/releases/tag/0.49.0) on GitHub.
diff --git a/docs/blog/2026-08-18-release-0-50.md b/docs/blog/2026-08-18-release-0-50.md
new file mode 100644
index 000000000..04703044e
--- /dev/null
+++ b/docs/blog/2026-08-18-release-0-50.md
@@ -0,0 +1,50 @@
+---
+date: '2026-08-18'
+title: 'Release 0.50.0'
+description: 'Safer test filenames, readable reports, easier replay of random test order and faster test runs.'
+coverAlt: 'Release 0.50.0 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.50.0 fixes a security problem with test filenames and cases where reports disagreed with the terminal summary. It also makes random test order easier to repeat and removes the interactive tutorial.
+
+## Test filenames cannot become commands
+
+The cleanup command run at the end of each test used to store the test-file path as Bash code. Bash read that code again when the test finished. A filename containing `$(...)` could therefore run a command. This matters when automated checks run tests from a branch whose filenames another contributor controls ([#1315](https://github.com/TypedDevs/bashunit/issues/1315)).
+
+Bash now treats the filename as text. These names also no longer prevent the test's checks from running and make it appear risky, the label used when a test may not have checked anything. Update projects that run tests from untrusted branches.
+
+## Repeat a random order with its seed
+
+A seed is the number used to reproduce a random test order. Passing `--seed` now selects that random order automatically. Copying the seed from a failing run is enough to repeat its order, unless you explicitly choose another order ([#1287](https://github.com/TypedDevs/bashunit/issues/1287)).
+
+```bash
+bashunit --seed 12345 tests/
+```
+
+Snapshots save expected test output for later comparison. Updating them also has a shorter spelling, `-u`:
+
+```bash
+bashunit -u --filter test_render_status tests/
+```
+
+## Reports agree with the run
+
+JSON and JUnit reports no longer contain stray terminal text when a parallel run selects no tests, finishes file cleanup, finds repeated function names or reports a command error. Reports also count a failed file setup or cleanup step once, matching the terminal summary ([#1295](https://github.com/TypedDevs/bashunit/issues/1295), [#1297](https://github.com/TypedDevs/bashunit/issues/1297), [#1299](https://github.com/TypedDevs/bashunit/issues/1299), [#1301](https://github.com/TypedDevs/bashunit/issues/1301)).
+
+Special characters in paths no longer break JUnit or Cobertura, the XML formats used for test and coverage reports. TAP reports sent directly to the terminal handle `#` in test descriptions. GitHub Actions error messages safely include special characters in titles and paths, and Markdown summaries handle failure messages that contain triple backticks. Benchmark files that fail to load or set up now make `bashunit bench` report failure.
+
+The recorded parallel reporting example runs about 1.9 times faster and uses about 1.8 times less processor time after removing repeated calls to base64, the tool used to encode report values. Startup saves about 9 ms in the recorded measurement because loading the display colours no longer starts an extra program for each colour ([#1289](https://github.com/TypedDevs/bashunit/issues/1289), [#1285](https://github.com/TypedDevs/bashunit/issues/1285)).
+
+## Tutorial removal
+
+`bashunit learn` has been removed. Calling it explains the removal and points to [the documentation](/installation). The guides provide learning material and examples; the runner no longer includes the interactive tutorial.
+
+---
+
+See the full changelog for [0.50.0](https://github.com/TypedDevs/bashunit/releases/tag/0.50.0) on GitHub.
diff --git a/docs/blog/2026-08-22-release-0-50-1.md b/docs/blog/2026-08-22-release-0-50-1.md
new file mode 100644
index 000000000..70b7d8c55
--- /dev/null
+++ b/docs/blog/2026-08-22-release-0-50-1.md
@@ -0,0 +1,36 @@
+---
+date: '2026-08-22'
+title: 'Release 0.50.1'
+description: 'Cleanup runs after setup failures, time limits and interruptions, and invalid test settings correctly fail the run.'
+coverAlt: 'Release 0.50.1 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.50.1 focuses on cleanup. Setup and cleanup functions now release resources after failures, early stops, invalid test settings and interrupted runs. This works both when tests run one after another and when they run at the same time.
+
+## Shared files stay available until their tests finish
+
+With `--parallel`, `tear_down_after_script` now runs after all tests in its own file finish. A file or other resource created by `set_up_before_script` therefore stays available while those tests use it ([#1320](https://github.com/TypedDevs/bashunit/issues/1320)).
+
+The same cleanup function runs when setup fails, when `--stop-on-failure` ends a sequential run, or when invalid `@timeout` or `@retry` settings stop a test file. Benchmark files also run their cleanup function before stopping because of an invalid test setting ([#1318](https://github.com/TypedDevs/bashunit/issues/1318), [#1321](https://github.com/TypedDevs/bashunit/issues/1321), [#1322](https://github.com/TypedDevs/bashunit/issues/1322), [#1329](https://github.com/TypedDevs/bashunit/issues/1329)).
+
+## Cleanup after interruption
+
+Ctrl-C runs the interrupted file's cleanup function. When tests run at the same time, it also runs the cleanup for a test still in progress. The shell process responsible for those tests handles the interruption. A second Ctrl-C ends a sequential run even if cleanup does not finish ([#1323](https://github.com/TypedDevs/bashunit/issues/1323), [#1331](https://github.com/TypedDevs/bashunit/issues/1331)).
+
+A test stopped by `--test-timeout` gets a chance to run `tear_down`. There is still a time limit on cleanup before bashunit forcibly stops it, so a cleanup function that hangs cannot keep the run open indefinitely ([#1324](https://github.com/TypedDevs/bashunit/issues/1324)).
+
+Temporary files created outside a test function with `bashunit::temp_file` are also removed when their test file fails to load.
+
+## Errors reach automated checks
+
+Invalid timeout or retry settings now fail the overall parallel run, even when another file passes. Previously one test process could print the error while the main process still reported success ([#1335](https://github.com/TypedDevs/bashunit/issues/1335)). Cleanup and the final result now reflect what happened in each file.
+
+---
+
+See the full changelog for [0.50.1](https://github.com/TypedDevs/bashunit/releases/tag/0.50.1) on GitHub.
diff --git a/docs/blog/2026-09-18-release-0-51.md b/docs/blog/2026-09-18-release-0-51.md
new file mode 100644
index 000000000..adf0d0ea2
--- /dev/null
+++ b/docs/blog/2026-09-18-release-0-51.md
@@ -0,0 +1,40 @@
+---
+date: '2026-09-18'
+title: 'Release 0.51.0'
+description: 'Faster setup and cleanup, more accurate coverage and fixes for test names and time limits.'
+coverAlt: 'Release 0.51.0 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.51.0 reduces the work around each test and still supports Bash 3.0. It also corrects coverage, which shows the code your tests ran, for statements spread across several lines. Similar test names and cleanup after a time limit work more reliably too.
+
+## Faster setup, cleanup and test counting
+
+Tests that define `set_up` or `tear_down` no longer create and remove extra output files for each call. In a recorded measurement on a Mac with an ARM processor and Bash 3.2, a test using these functions fell from 14.1 ms to 5.4 ms, about 2.6 times faster ([#1345](https://github.com/TypedDevs/bashunit/issues/1345)).
+
+bashunit now looks up a test's file and line only when needed instead of starting a separate shell for every test run in sequence. In another measurement on the same platform, a 500-test file fell from 2.75 seconds to 2.40 seconds ([#1346](https://github.com/TypedDevs/bashunit/issues/1346)).
+
+To show the number of tests before a run, bashunit reuses information it already found while reading test functions. Files whose test count needs more work are still loaded as before. These include files that supply several sets of test inputs, contain text spread across lines, or create functions while they run. Across the repository's 241-file measurement, counting fell from 1.72 seconds to 1.51 seconds ([#1347](https://github.com/TypedDevs/bashunit/issues/1347)).
+
+When Bash cannot read the clock directly through `EPOCHREALTIME`, parallel runs also stop checking which clock to use for every test. The 500-test check now starts Perl twice instead of 502 times.
+
+## More accurate coverage
+
+Coverage counts every line of arrays, quoted text, blocks of text written with `<<`, and commands continued onto another line. Previously an array spread across several lines could look partly untested, and on Bash 3.x could look entirely untested. Running an outer command also no longer marks commands inside `$(...)` or `<(...)` as tested unless they ran ([#1338](https://github.com/TypedDevs/bashunit/issues/1338)).
+
+## Test names and cleanup fixes
+
+`test_a` and `test_ab` are no longer mistaken for the same function defined twice. An `--exclude-filter` value containing a comma stays one test name; repeat the option to exclude several names. The environment setting `BASHUNIT_EXCLUDE_FILTER` still uses commas to separate names.
+
+Tests that reach their time limit get a limited chance to finish `tear_down`. bashunit waits for the test itself to exit rather than assuming a short fixed delay is always enough. Parallel runs also stop displaying Bash's warning about grouping running processes as though a test file had printed it ([#1344](https://github.com/TypedDevs/bashunit/issues/1344)).
+
+Helpers can now choose faster code for a newer Bash when loaded, while keeping the existing code for Bash 3.0. The minimum supported version stays the same.
+
+---
+
+See the full changelog for [0.51.0](https://github.com/TypedDevs/bashunit/releases/tag/0.51.0) on GitHub.
diff --git a/docs/blog/2026-10-11-release-0-52.md b/docs/blog/2026-10-11-release-0-52.md
new file mode 100644
index 000000000..30aa00997
--- /dev/null
+++ b/docs/blog/2026-10-11-release-0-52.md
@@ -0,0 +1,36 @@
+---
+date: '2026-10-11'
+title: 'Release 0.52.0'
+description: 'Less work when supplying test inputs and writing reports, reliable parallel results and cleanup, and shared contributor instructions.'
+coverAlt: 'Release 0.52.0 highlights'
+
+aside: false
+---
+
+# {{ $frontmatter.title }}
+
+<time>{{ $formatDate($frontmatter.date) }}</time>
+
+Release 0.52.0 does less work when reading test inputs and collecting reports. It also keeps separate test results and cleanup working correctly when tests run at the same time. Quoted inputs, empty values and report contents keep their existing behavior.
+
+## Less work between tests
+
+Parallel reports no longer start base64, the tool used to encode values, separately for each result. Data providers, functions that supply several sets of inputs to a test, also avoid calling it for each argument. Quotes and empty values are preserved. Reading those arguments still happens in a separate shell so it cannot change the main shell's state ([#1371](https://github.com/TypedDevs/bashunit/issues/1371), [#1370](https://github.com/TypedDevs/bashunit/issues/1370)).
+
+JSON reports avoid starting an extra program for each ordinary filename, test name or empty message. Special characters are still written safely. When running files one after another, bashunit also reuses saved information about test inputs and settings for unchanged files up to 512 bytes. It checks that the file still has the same contents before reusing that information ([#1369](https://github.com/TypedDevs/bashunit/issues/1369), [#1372](https://github.com/TypedDevs/bashunit/issues/1372)).
+
+## Each set of inputs keeps its own result
+
+When test cases run at the same time, each manages its own temporary files, even if their arguments are identical. Cases that run one after another inside a parallel run also keep separate results, so a later passing case cannot hide an earlier failure ([#1379](https://github.com/TypedDevs/bashunit/issues/1379), [#1377](https://github.com/TypedDevs/bashunit/issues/1377)).
+
+A request to stop a test that arrives late no longer interrupts `tear_down`. Cleanup still has a time limit if it hangs. Sending parallel output to another command also no longer waits for the unused progress indicator to finish sleeping ([#1376](https://github.com/TypedDevs/bashunit/issues/1376), [#1368](https://github.com/TypedDevs/bashunit/issues/1368)).
+
+## Shared contributor setup
+
+Contributor instructions now have one shared source in agnostic-ai, which creates the files read by Claude Code and Codex. The project installs a fixed version of this tool through npm. The shared instructions decide which guidance and model to use for each task. Contributors edit `.agnostic-ai/` and then recreate the tool-specific files ([#1387](https://github.com/TypedDevs/bashunit/pull/1387)).
+
+The release script now updates and commits the project version in both npm lockfiles, which record installed dependencies. It restores those files if local release changes need to be undone. Automated checks for supported Bash versions read shell code without treating generated instructions as code, and the Linux installer checks now install the build tool they require.
+
+---
+
+See the full changelog for [0.52.0](https://github.com/TypedDevs/bashunit/releases/tag/0.52.0) on GitHub.